Postoperative Spectral Head CT Exposes a Three-Way Trade-off in Virtual Monoenergetic Imaging
Postoperative spectral head CT can be reconstructed into multiple virtual monoenergetic images, but the best energy level is not uniform across image-quality goals. In this two-center reader study, 93 patients contributed 120 postoperative examinations, and the results separated cleanly by what was being measured: contrast, noise, or the radiologist’s visual impression at the bedside.
The central message is blunt. No single VMI energy level optimized every relevant image-quality property at once.
That matters in the postoperative head, where metallic artifact, air, blood products, and surgical-bed changes can crowd the image and make different viewing priorities compete with one another. A reconstruction that sharpens one feature may soften another. Here, the data laid out those trade-offs directly rather than pretending they disappear with a single preferred preset.
60 keV Maximized Contrast, While 100 keV Minimized Noise
Investigators compared conventional images with VMIs reconstructed at 60, 70, 80, 90, and 100 keV. Objective image metrics were analyzed alongside independent ratings from two neuroradiologists, with mixed-effects models used to account for the hierarchical structure of the data. Among the objective measures, contrast-to-noise ratio was highest at 60 keV, reaching 1.62 ± 0.39. At the other end of the spectrum, image noise was lowest at 100 keV, at 4.27 ± 0.49 HU.
Those findings fit the physics of monoenergetic reconstruction. Lower-energy VMIs tend to preserve or amplify iodine-related contrast, which can sharpen conspicuity at the surgical margin and around enhancing structures. Higher-energy images generally blunt beam-hardening and other spectral artifacts while calming the graininess that can make postoperative scans difficult to read. In a head CT performed after surgery, those goals are often in tension. The study did not flatten that tension; it quantified it.
Reader Preference Concentrated at 70 keV
The most clinically persuasive result came from the neuroradiologists’ subjective ratings. The 70-keV VMI earned the highest scores for overall image quality, operative-bed parenchymal conspicuity, and diagnostic acceptability. In practical terms, that suggests a middle-energy reconstruction offered the most balanced appearance for routine review, even though it did not maximize the objective CNR observed at 60 keV.
One sentence captures the pattern: the most contrast-rich image was not the one readers liked best.
This is a familiar divergence in radiology, and postoperative neuroimaging is especially sensitive to it. Human readers do not evaluate only one metric; they integrate edge definition, tissue conspicuity, artifact suppression, and confidence in the surgical bed within a single glance. A reconstruction can register well on one numerical measure and still feel suboptimal at the workstation. The 70-keV result suggests the visual sweet spot may sit between the contrast-maximizing low-energy setting and the noise-suppressing high-energy setting.
Artifact Suppression Favored the Highest Energy Setting
The 100-keV VMI produced the highest artifact-suppression score. That result is consistent with the expected behavior of higher-energy reconstructions, which often reduce the severity of streaks and other spectral distortions that can obscure anatomy after cranial surgery. In a postoperative scan, that can be clinically useful when hardware, blood products, or altered skull contours interfere with interpretation.
Still, artifact suppression did not translate into the best overall reader impression. The same reconstruction that cleaned up the image also delivered the highest noise suppression, but readers favored 70 keV for overall quality and acceptability. The study therefore points to a workflow problem rather than a single best answer: a radiologist deciding what matters most in a specific scan may reasonably prefer a different energy level depending on whether the main concern is surgical-bed assessment, artifact burden, or simple image smoothness.
Agreement Between Readers Was Strong Across the Spectrum
Inter-reader agreement was high throughout the analysis, with quadratic-weighted kappa values ranging from 0.859 to 0.986. In a qualitative reader study, that degree of concordance supports the reliability of the subjective scoring pattern. Two neuroradiologists reached broadly similar judgments about which reconstructions looked best for routine postoperative review, which strengthens the practical meaning of the 70-keV preference.
High agreement also reduces the chance that the findings were driven by one reader’s idiosyncratic visual threshold. The consistency matters because reader-rated image quality is not a secondary ornament in postoperative neuroimaging; it often determines whether a scan feels comfortably diagnostic or merely technically adequate.
Why the Energy Choice Will Likely Stay Task-Specific
The study’s design supports a task-specific interpretation of VMI rather than a single universal setting. Low-energy reconstructions can raise contrast, which is appealing when subtle parenchymal differences or operative-bed detail must stand out. High-energy reconstructions can lower noise and mute artifacts, which helps when streaks or granular texture threaten interpretability. Midrange energy may offer the best compromise when the question is routine postoperative review rather than one narrowly defined imaging target.
That compromise is the real clinical lesson embedded in the numbers. A postoperative head CT is not a static object; it is a difficult visual environment in which artifact, noise, and contrast compete for attention. The image that looks best to the eye may not be the one with the best quantitative contrast ratio, and the reconstruction that looks smoothest may not be the one that most clearly defines the operative bed.
Diagnostic-performance validation still has to happen before these findings can be folded into a task-specific workflow. This study compared image-quality endpoints, not hard diagnostic outcomes such as lesion detection, complication recognition, or management changes. The authors also studied a postoperative cohort from two centers, so the next question is not whether VMI trade-offs exist — they do — but how those trade-offs perform when the scan has to answer a concrete clinical question at the point of care.